git commit -m "looking for someone who gets it"

Your GitHub profile is a better personality test than any questionnaire. The languages you reach for, the repos you star at midnight, the side projects you start (and the ones you actually finish) - that is who you are.

Try the demo

How it works for devs

Connect your repos. GitHub, GitLab, Bitbucket - we pull your public languages, topics, contribution patterns, and starred repos. Your code stays yours. We only read metadata.

The matching engine understands code at multiple levels:

A repo uses languages, covers topics, and shows contribution cadence
Matching happens across the stack - shared languages, shared topics, similar build patterns
Repos per language is the engagement signal - one Python repo is presence; 15 Python repos is identity
If we have repo-level data but no explicit topic affinity, we compute it from your repository metadata

Your stack is your signal

Two devs both have "Python" on their profile. One has a single tutorial repo. The other has 23 repos, 400+ commits across data science, automation, and API projects. The algorithm knows the difference.

EngagementFactor uses repos-per-language as depth. The geometric mean ensures both people need meaningful investment for a high score.

Niche languages and topics hit different

Mainstream

JavaScript

95% of devs → NicheWeight 0.05

Niche

Haskell

8% of devs → NicheWeight 0.92

The niche proxy for code uses language and topic frequency across the dev population. Rare languages, obscure frameworks, and niche topics contribute the most signal.

What gets scored

Shared languages weighted by rarity (Rust scores higher than JavaScript)

Repos per language - depth of investment, not just "used it once"

Shared topics - machine-learning, cli-tools, game-dev, embedded-systems

Contribution cadence - daily committer vs weekend warrior vs burst contributor

Starred repo overlap - the repos you bookmark say as much as the ones you write

Shared avoidances - neither of you touching PHP despite it being everywhere? That is a shared value

Builder mindset signals - README quality, test coverage patterns, documentation habits

README.md for your love life

// match_report.md

- Shared languages: Rust (10+ repos each), Go (5+ repos each)
- Shared topics: wasm, compilers, embedded, systems-programming
- Cadence overlap: 74%
- Shared avoidance: PHP (0 repos, both)
- Niche score: HIGH

No "you both like coding." The diff.

Pair programming for life

This is not about finding someone to do LeetCode with (unless that is your thing). It is about finding someone whose brain works like yours - someone who stars the same obscure projects, picks the same tools for the same reasons, and understands why you spent a weekend rewriting a perfectly working function.

Integrations

GitHub

Public repos, languages, topics, stars, contribution graph

GitLab

Projects, languages, activity

Bitbucket

Repositories, languages

All opt-in. All revocable. Private repos are never accessed.

Affinity Atlas is in development

The codebase is a solo project by a developer and data scientist who thinks dating apps should use better data. If you want to follow along, contribute ideas, or just share your most controversial language opinion - get in touch.

Stay in the loop